Meetings, status, alignment, rework - the invisible cost of keeping teams in sync. AI's productivity gains are being absorbed by it before they reach the bottom line.
Most leadership teams know the coordination tax exists but underestimate its scale. It's invisible on the P&L. It grows exponentially as the org scales. And it eats AI productivity gains before they show up in margin.
The first place AI changes the org isn't the worker - it's the coordination layer the worker sits inside.
With n teams in an org, the number of possible coordination paths grows roughly as n². Add a tenth team and you don't add 10% more coordination - you add 20% more coordination paths. This is why org growth feels like falling off a productivity cliff somewhere around 50 to 200 people.
The coordination tax shows up as meetings, status updates, alignment overhead, rework after teams discover they were building incompatible things. None of it is visible on a financial statement. All of it is real time spent.
AI promises to absorb this by maintaining shared context across teams continuously - turning meetings into AI-readable artifacts, status into auto-generated rollups, alignment into a live shared representation. But only if you wire it in as a coordination layer. Bolt it on as a tool and the tax keeps growing while your AI line item also grows.
When the world's best engineers paid the coordination tax in cash.
September 23, 1999. NASA's Mars Climate Orbiter approached Mars after a 286-day journey. Routine final manoeuvre. The signal went dead. The $125 million spacecraft had disintegrated in the atmosphere.
The cause: Lockheed Martin did the math in pound-force. NASA's Jet Propulsion Laboratory assumed Newtons. The numbers were off by a factor of 4.45. Over millions of miles, the tiny errors compounded. Both teams were excellent. Each performed their role flawlessly. They just weren't coordinated.
Every large organisation runs Mars Climate Orbiter missions, every day, at small scale. Two expert teams, each performing well, building incompatible things because a coordination assumption never got reconciled. NASA paid $125 million for theirs. Most companies pay in margin and velocity.
At 50 people, this is maybe 20% of time. At 5,000, it's often 60–70% - and almost nobody has it on a P&L line. AI can either absorb it or amplify it, depending entirely on whether you wire AI in as coordination or as automation.
Big enterprise rolls out Copilot to 10,000 employees. Each employee is individually more productive. The hours saved go into… more meetings. More status updates. More cross-team alignment about what each AI-augmented team is doing.
The coordination tax absorbs the productivity gain. The CFO can't explain why the AI spend hasn't shown up in margin. Because AI was wired in as automation, not as coordination.
AI-native firms wire AI in differently: every Slack message, every meeting transcript, every document becomes part of a continuously updated shared context that the org's AI systems read and reason over.
The result: dramatically lower coordination tax, even at scale. They run with smaller teams that don't need constant alignment meetings because the context is always in sync. This is where the real productivity gains live.
Pick a function. Slide right to represent the org scaling. Tasks migrate from "real work" into the coordination-tax buckets. What's left at the top is the proportion of time spent on what customers pay for.
Pick whichever is closest to your work. The mechanism is the same for any knowledge role.
Bolt AI on and the tax eats the gains. Wire AI in as coordination and the tax goes down.
An advisory engagement that redesigns the operating model around AI as a coordination capability - so initiatives compound into a shared capability architecture instead of fragmenting into disconnected automations the coordination tax then absorbs.